{"id":15020960,"url":"https://github.com/swanhubx/swanlab","last_synced_at":"2025-05-13T23:09:43.313Z","repository":{"id":213707646,"uuid":"722915433","full_name":"SwanHubX/SwanLab","owner":"SwanHubX","description":"⚡️SwanLab - an open-source, modern-design AI training tracking and visualization tool. Supports Cloud / Self-hosted use. Integrated with PyTorch / Transformers / LLaMA Factory / Swift / Ultralytics / veRL / MMEngine / Keras etc.","archived":false,"fork":false,"pushed_at":"2025-05-07T14:00:53.000Z","size":43686,"stargazers_count":1470,"open_issues_count":43,"forks_count":96,"subscribers_count":5,"default_branch":"main","last_synced_at":"2025-05-07T14:37:12.187Z","etag":null,"topics":["data-science","deep-learning","jax","logging","machine-learning","mlops","model-versioning","python","pytorch","tensorboard","tensorflow","tracking","transformers","visualization"],"latest_commit_sha":null,"homepage":"https://swanlab.cn","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/SwanHubX.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":"CODE_OF_CONDUCT.md","threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":"SECURITY.md","support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2023-11-24T08:54:45.000Z","updated_at":"2025-05-07T13:19:57.000Z","dependencies_parsed_at":"2023-12-26T16:40:58.194Z","dependency_job_id":"ae25129a-2fca-4397-8c87-d0155a5fde7e","html_url":"https://github.com/SwanHubX/SwanLab","commit_stats":{"total_commits":428,"total_committers":18,"mean_commits":23.77777777777778,"dds":0.6401869158878505,"last_synced_commit":"63bd0971bbb6ccd09f620e68e6b271a5d0ca9561"},"previous_names":["swanhubx/swanlab"],"tags_count":81,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/SwanHubX%2FSwanLab","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/SwanHubX%2FSwanLab/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/SwanHubX%2FSwanLab/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/SwanHubX%2FSwanLab/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/SwanHubX","download_url":"https://codeload.github.com/SwanHubX/SwanLab/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":253950587,"owners_count":21989373,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["data-science","deep-learning","jax","logging","machine-learning","mlops","model-versioning","python","pytorch","tensorboard","tensorflow","tracking","transformers","visualization"],"created_at":"2024-09-24T19:55:56.731Z","updated_at":"2025-05-13T23:09:38.262Z","avatar_url":"https://github.com/SwanHubX.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003cdiv align=\"center\"\u003e\n\n\u003cpicture\u003e\n  \u003csource media=\"(prefers-color-scheme: dark)\" srcset=\"readme_files/swanlab-logo-single-dark.svg\"\u003e\n  \u003csource media=\"(prefers-color-scheme: light)\" srcset=\"readme_files/swanlab-logo-single.svg\"\u003e\n  \u003cimg alt=\"SwanLab\" src=\"readme_files/swanlab-logo-single.svg\" width=\"70\" height=\"70\"\u003e\n\u003c/picture\u003e\n\n\u003ch1\u003eSwanLab\u003c/h1\u003e\n\n一个开源、现代化设计的深度学习训练跟踪与可视化工具  \n同时支持云端/离线使用，适配30+主流框架，与你的实验代码轻松集成\n\n\u003ca href=\"https://swanlab.cn\"\u003e🔥SwanLab 在线版\u003c/a\u003e · \u003ca href=\"https://docs.swanlab.cn\"\u003e📃 文档\u003c/a\u003e · \u003ca href=\"https://github.com/swanhubx/swanlab/issues\"\u003e报告问题\u003c/a\u003e · \u003ca href=\"https://geektechstudio.feishu.cn/share/base/form/shrcnyBlK8OMD0eweoFcc2SvWKc\"\u003e建议反馈\u003c/a\u003e · \u003ca href=\"https://docs.swanlab.cn/zh/guide_cloud/general/changelog.html\"\u003e更新日志\u003c/a\u003e\n\n[![][release-shield]][release-link]\n[![][dockerhub-shield]][dockerhub-link]\n[![][github-stars-shield]][github-stars-link]\n[![][github-issues-shield]][github-issues-shield-link]\n[![][github-contributors-shield]][github-contributors-link]\n[![][license-shield]][license-shield-link]  \n[![][tracking-swanlab-shield]][tracking-swanlab-shield-link]\n[![][last-commit-shield]][last-commit-shield-link]\n[![][pypi-version-shield]][pypi-version-shield-link]\n[![][wechat-shield]][wechat-shield-link]\n[![][pypi-downloads-shield]][pypi-downloads-shield-link]\n[![][colab-shield]][colab-shield-link]\n\n\n![](readme_files/swanlab-overview.png)\n\n中文 / [English](README_EN.md) / [日本語](README_JP.md) / [Русский](README_RU.md)\n\n👋 加入我们的[微信群](https://docs.swanlab.cn/zh/guide_cloud/community/online-support.html)\n\n\u003ca href=\"https://hellogithub.com/repository/b442a9fa270e4ccb8847c9ee3445e41b\" target=\"_blank\"\u003e\u003cimg src=\"https://abroad.hellogithub.com/v1/widgets/recommend.svg?rid=b442a9fa270e4ccb8847c9ee3445e41b\u0026claim_uid=Oh5UaGjfrblg0yZ\" alt=\"Featured｜HelloGitHub\" style=\"width: 250px; height: 54px;\" width=\"250\" height=\"54\" /\u003e\u003c/a\u003e\n\n\n\u003c/div\u003e\n\n\u003cbr/\u003e\n\n\n## 目录\n\n- [🌟 最近更新](#-最近更新)\n- [👋🏻 什么是SwanLab](#-什么是swanlab)\n- [📃 在线演示](#-在线演示)\n- [🏁 快速开始](#-快速开始)\n- [💻 自托管](#-自托管)\n- [🚗 框架集成](#-框架集成)\n- [🔌 插件](#-插件)\n- [🎮 硬件记录](#-硬件记录)\n- [🆚 与熟悉的工具的比较](#-与熟悉的工具的比较)\n- [👥 社区](#-社区)\n- [📃 协议](#-协议)\n\n\u003cbr/\u003e\n\n\n## 🌟 最近更新\n\n- 2025.04.11：支持折线图**局部区域选取**；支持全局当前仪表盘折线图的step范围；\n\n- 2025.04.08：支持**swanlab.Molecule**数据类型，支持记录与可视化生物化学分子数据；支持保存表格视图中的排序、筛选、列顺序变化状态；\n\n- 2025.04.07：我们与 [EvalScope](https://github.com/ModelScope/EvalScope) 完成了联合集成，现在你可以在EvalScope中使用SwanLab来**评估大模型性能**；\n\n- 2025.03.30：支持**swanlab.Settings**方法，支持更精细化的实验行为控制；支持**寒武纪MLU**硬件监控；支持 [Slack通知](https://docs.swanlab.cn/plugin/notification-slack.html)、[Discord通知](https://docs.swanlab.cn/plugin/notification-discord.html)；\n\n- 2025.03.21：🎉🤗HuggingFace Transformers已正式集成SwanLab（\u003e=4.50.0版本），[#36433](https://github.com/huggingface/transformers/pull/36433)；新增 **Object3D图表** ，支持记录与可视化三维点云，[文档](https://docs.swanlab.cn/api/py-object3d.html)；硬件监控支持了 GPU显存（MB）、磁盘利用率、网络上下行 的记录；\n\n- 2025.03.12：🎉🎉SwanLab**私有化部署版**现已发布！！[🔗部署文档](https://docs.swanlab.cn/guide_cloud/self_host/docker-deploy.html)；SwanLab 已支持插件扩展，如 [邮件通知](https://docs.swanlab.cn/plugin/notification-email.html)、[飞书通知](https://docs.swanlab.cn/plugin/notification-lark.html)\n\n- 2025.03.09：支持**实验侧边栏拉宽**；新增外显 Git代码 按钮；新增 **sync_mlflow** 功能，支持与mlflow框架同步实验跟踪；\n\n- 2025.03.06：我们与 [DiffSynth Studio](https://github.com/modelscope/diffsynth-studio) 完成了联合集成，现在你可以在DiffSynth Studio中使用SwanLab来**跟踪和可视化Diffusion模型文生图/视频实验**，[使用指引](https://docs.swanlab.cn/guide_cloud/integration/integration-diffsynth-studio.html)；\n\n- 2025.03.04：新增 **MLFlow转换** 功能，支持将MLFlow实验转换为SwanLab实验，[使用指引](https://docs.swanlab.cn/guide_cloud/integration/integration-mlflow.html)；\n\n\n\u003cdetails\u003e\u003csummary\u003e完整更新日志\u003c/summary\u003e\n\n- 2025.03.01：新增 **移动实验** 功能，现在可以将实验移动到不同组织的不同项目下了；\n\n- 2025.02.24：我们与 [EasyR1](https://github.com/hiyouga/EasyR1) 完成了联合集成，现在你可以在EasyR1中使用SwanLab来**跟踪和可视化多模态大模型强化学习实验**，[使用指引](https://docs.swanlab.cn/guide_cloud/integration/integration-easyr1.html)\n\n- 2025.02.18：我们与 [Swift](https://github.com/modelscope/ms-swift) 完成了联合集成，现在你可以在Swift的CLI/WebUI中使用SwanLab来**跟踪和可视化大模型微调实验**，[使用指引](https://docs.swanlab.cn/guide_cloud/integration/integration-swift.html)。\n\n- 2025.02.16：新增 **图表移动分组、创建分组** 功能。\n\n- 2025.02.09：我们与 [veRL](https://github.com/volcengine/verl) 完成了联合集成，现在你可以在veRL中使用SwanLab来**跟踪和可视化大模型强化学习实验**，[使用指引](https://docs.swanlab.cn/guide_cloud/integration/integration-verl.html)。\n\n- 2025.02.05：`swanlab.log`支持嵌套字典 [#812](https://github.com/SwanHubX/SwanLab/pull/812)，适配Jax框架特性；支持`name`与`notes`参数；\n\n- 2025.01.22：新增`sync_tensorboardX`与`sync_tensorboard_torch`功能，支持与此两种TensorBoard框架同步实验跟踪；\n\n- 2025.01.17：新增`sync_wandb`功能，[文档](https://docs.swanlab.cn/guide_cloud/integration/integration-wandb.html)，支持与Weights \u0026 Biases实验跟踪同步；大幅改进了日志渲染性能\n\n- 2025.01.11：云端版大幅优化了项目表格的性能，并支持拖拽、排序、筛选等交互\n\n- 2025.01.01：新增折线图**持久化平滑**、折线图拖拽式改变大小，优化图表浏览体验\n\n- 2024.12.22：我们与 [LLaMA Factory](https://github.com/hiyouga/LLaMA-Factory) 完成了联合集成，现在你可以在LLaMA Factory中使用SwanLab来**跟踪和可视化大模型微调实验**，[使用指引](https://github.com/hiyouga/LLaMA-Factory?tab=readme-ov-file#use-swanlab-logger)。\n\n- 2024.12.15：**硬件监控（0.4.0）** 功能上线，支持CPU、NPU（Ascend）、GPU（Nvidia）的系统级信息记录与监控。\n\n- 2024.12.06：新增对[LightGBM](https://docs.swanlab.cn/guide_cloud/integration/integration-lightgbm.html)、[XGBoost](https://docs.swanlab.cn/guide_cloud/integration/integration-xgboost.html)的集成；提高了对日志记录单行长度的限制。\n\n- 2024.11.26：环境选项卡-硬件部分支持识别**华为昇腾NPU**与**鲲鹏CPU**；云厂商部分支持识别青云**基石智算**。\n\n\u003c/details\u003e\n\n\u003cbr\u003e\n\n## 👋🏻 什么是SwanLab\n\nSwanLab 是一款开源、轻量的 AI 模型训练跟踪与可视化工具，提供了一个跟踪、记录、比较、和协作实验的平台。\n\nSwanLab 面向人工智能研究者，设计了友好的Python API 和漂亮的UI界面，并提供**训练可视化、自动日志记录、超参数记录、实验对比、多人协同**等功能。在SwanLab上，研究者能基于直观的可视化图表发现训练问题，对比多个实验找到研究灵感，并通过**在线网页**的分享与基于组织的**多人协同训练**，打破团队沟通的壁垒，提高组织训练效率。\n\nhttps://github.com/user-attachments/assets/7965fec4-c8b0-4956-803d-dbf177b44f54\n\n以下是其核心特性列表：\n\n**1. 📊 实验指标与超参数跟踪**: 极简的代码嵌入您的机器学习 pipeline，跟踪记录训练关键指标\n\n- ☁️ 支持**云端**使用（类似Weights \u0026 Biases），随时随地查看训练进展。[手机看实验的方法](https://docs.swanlab.cn/guide_cloud/general/app.html)\n\n- 📝 支持**超参数记录**、**指标总结**、**表格分析**\n\n- 🌸 **可视化训练过程**: 通过UI界面对实验跟踪数据进行可视化，可以让训练师直观地看到实验每一步的结果，分析指标走势，判断哪些变化导致了模型效果的提升，从而整体性地提升模型迭代效率。\n\n- **支持的元数据类型**：标量指标、图像、音频、文本、3D点云、生物化学分子...\n\n- **支持的图表类型**：折线图、媒体图（图像、音频、文本）、3D点云、生物化学分子...\n\n![swanlab-table](readme_files/molecule.gif)\n\n- **后台自动记录**：日志logging、硬件环境、Git 仓库、Python 环境、Python 库列表、项目运行目录\n\n**2. ⚡️ 全面的框架集成**: PyTorch、🤗HuggingFace Transformers、PyTorch Lightning、🦙LLaMA Factory、MMDetection、Ultralytics、PaddleDetetion、LightGBM、XGBoost、Keras、Tensorboard、Weights\u0026Biases、OpenAI、Swift、XTuner、Stable Baseline3、Hydra 在内的 **30+** 框架\n\n![](readme_files/integrations.png)\n\n**3. 💻 硬件监控**: 支持实时记录与监控CPU、NPU（**昇腾Ascend**）、GPU（**英伟达Nvidia**）、MLU（**寒武纪MLU**）、内存的系统级硬件指标\n\n**4. 📦 实验管理**: 通过专为训练场景设计的集中式仪表板，通过整体视图速览全局，快速管理多个项目与实验\n\n**5. 🆚 比较结果**: 通过在线表格与对比图表比较不同实验的超参数和结果，挖掘迭代灵感\n\n![](readme_files/swanlab-table.png)\n\n**6. 👥 在线协作**: 您可以与团队进行协作式训练，支持将实验实时同步在一个项目下，您可以在线查看团队的训练记录，基于结果发表看法与建议\n\n**7. ✉️ 分享结果**: 复制和发送持久的 URL 来共享每个实验，方便地发送给伙伴，或嵌入到在线笔记中\n\n**8. 💻 支持自托管**: 支持离线环境使用，自托管的社区版同样可以查看仪表盘与管理实验，[使用攻略](#-自托管)\n\n**9. 🔌 插件拓展**: 支持通过插件拓展SwanLab的使用场景，比如 [飞书通知](https://docs.swanlab.cn/plugin/notification-lark.html)、[Slack通知](https://docs.swanlab.cn/plugin/notification-slack.html)、[CSV记录器](https://docs.swanlab.cn/plugin/writer-csv.html)等\n\n\u003e \\[!IMPORTANT]\n\u003e\n\u003e **收藏项目**，你将从 GitHub 上无延迟地接收所有发布通知～ ⭐️\n\n![star-us](readme_files/star-us.png)\n\n\u003cbr\u003e\n\n## 📃 在线演示\n\n来看看 SwanLab 的在线演示：\n\n| [ResNet50 猫狗分类][demo-cats-dogs] | [Yolov8-COCO128 目标检测][demo-yolo] |\n| :--------: | :--------: |\n| [![][demo-cats-dogs-image]][demo-cats-dogs] | [![][demo-yolo-image]][demo-yolo] |\n| 跟踪一个简单的 ResNet50 模型在猫狗数据集上训练的图像分类任务。 | 使用 Yolov8 在 COCO128 数据集上进行目标检测任务，跟踪训练超参数和指标。 |\n\n| [Qwen2 指令微调][demo-qwen2-sft] | [LSTM Google 股票预测][demo-google-stock] |\n| :--------: | :--------: |\n| [![][demo-qwen2-sft-image]][demo-qwen2-sft] | [![][demo-google-stock-image]][demo-google-stock] |\n| 跟踪 Qwen2 大语言模型的指令微调训练，完成简单的指令遵循。 | 使用简单的 LSTM 模型在 Google 股价数据集上训练，实现对未来股价的预测。 |\n\n| [ResNeXt101 音频分类][demo-audio-classification] | [Qwen2-VL COCO数据集微调][demo-qwen2-vl] |\n| :--------: | :--------: |\n| [![][demo-audio-classification-image]][demo-audio-classification] | [![][demo-qwen2-vl-image]][demo-qwen2-vl] |\n| 从ResNet到ResNeXt在音频分类任务上的渐进式实验过程 | 基于Qwen2-VL多模态大模型，在COCO2014数据集上进行Lora微调。 |\n\n| [EasyR1 多模态LLM RL训练][demo-easyr1-rl] | [Qwen2.5-0.5B GRPO训练][demo-qwen2-grpo] |\n| :--------: | :--------: |\n| [![][demo-easyr1-rl-image]][demo-easyr1-rl] | [![][demo-qwen2-grpo-image]][demo-qwen2-grpo] |\n| 使用EasyR1框架进行多模态LLM RL训练 | 基于Qwen2.5-0.5B模型在GSM8k数据集上进行GRPO训练 |\n\n[更多案例](https://docs.swanlab.cn/zh/examples/mnist.html)\n\n\u003cbr\u003e\n\n## 🏁 快速开始\n\n### 1.安装\n\n```bash\npip install swanlab\n```\n\n\u003cdetails\u003e\u003csummary\u003e源码安装\u003c/summary\u003e\n\n如果你想体验最新的特性，可以使用源码安装。\n\n```bash\n# 方式一\ngit clone https://github.com/SwanHubX/SwanLab.git\npip install -e .\n\n# 方式二\npip install git+https://github.com/SwanHubX/SwanLab.git\n```\n\n\u003c/details\u003e\n\n\u003cdetails\u003e\u003csummary\u003e离线看板拓展安装\u003c/summary\u003e\n\n[离线看板文档](https://docs.swanlab.cn/guide_cloud/self_host/offline-board.html)\n\n```bash\npip install 'swanlab[dashboard]'\n```\n\n\u003c/details\u003e\n\n\n### 2.登录并获取 API Key\n\n1. 免费[注册账号](https://swanlab.cn)\n\n2. 登录账号，在用户设置 \u003e [API Key](https://swanlab.cn/settings) 里复制您的 API Key\n\n3. 打开终端，输入：\n\n```bash\nswanlab login\n```\n\n出现提示时，输入您的 API Key，按下回车，完成登陆。\n\n### 3.将 SwanLab 与你的代码集成\n\n```python\nimport swanlab\n\n# 初始化一个新的swanlab实验\nswanlab.init(\n    project=\"my-first-ml\",\n    config={'learning-rate': 0.003},\n)\n\n# 记录指标\nfor i in range(10):\n    swanlab.log({\"loss\": i, \"acc\": i})\n```\n\n大功告成！前往[SwanLab](https://swanlab.cn)查看你的第一个 SwanLab 实验。\n\n\u003cbr\u003e\n\n## 💻 自托管\n\n自托管社区版支持离线查看 SwanLab 仪表盘。\n\n![swanlab-docker](./readme_files/swanlab-docker.png)\n\n### 1. 使用Docker部署自托管版本\n\n详情请参考：[文档](https://docs.swanlab.cn/guide_cloud/self_host/docker-deploy.html)\n\n```bash\ngit clone https://github.com/SwanHubX/self-hosted.git\ncd self-hosted/docker\n```\n\n中国地区快速安装：\n\n```bash\n./install.sh\n```\n\n从DockerHub拉取镜像安装：\n\n```bash\n./install-dockerhub.sh\n```\n\n### 2. 将实验指定到自托管服务\n\n登录到自托管服务：\n\n```bash\nswanlab login --host http://localhost:8000\n```\n\n完成登录后，即可将实验记录到自托管服务。\n\n\n\u003cbr\u003e\n\n## 🚗 框架集成\n\n将你最喜欢的框架与 SwanLab 结合使用！  \n下面是我们已集成的框架列表，欢迎提交 [Issue](https://github.com/swanhubx/swanlab/issues) 来反馈你想要集成的框架。\n\n**基础框架**\n- [PyTorch](https://docs.swanlab.cn/guide_cloud/integration/integration-pytorch.html)\n- [MindSpore](https://docs.swanlab.cn/guide_cloud/integration/integration-ascend.html)\n- [Keras](https://docs.swanlab.cn/guide_cloud/integration/integration-keras.html)\n\n**专有/微调框架**\n- [PyTorch Lightning](https://docs.swanlab.cn/guide_cloud/integration/integration-pytorch-lightning.html)\n- [HuggingFace Transformers](https://docs.swanlab.cn/guide_cloud/integration/integration-huggingface-transformers.html)\n- [LLaMA Factory](https://docs.swanlab.cn/guide_cloud/integration/integration-llama-factory.html)\n- [Modelscope Swift](https://docs.swanlab.cn/guide_cloud/integration/integration-swift.html)\n- [DiffSynth Studio](https://docs.swanlab.cn/guide_cloud/integration/integration-diffsynth-studio.html)\n- [Sentence Transformers](https://docs.swanlab.cn/guide_cloud/integration/integration-sentence-transformers.html)\n- [OpenMind](https://modelers.cn/docs/zh/openmind-library/1.0.0/basic_tutorial/finetune/finetune_pt.html#%E8%AE%AD%E7%BB%83%E7%9B%91%E6%8E%A7)\n- [Torchtune](https://docs.swanlab.cn/guide_cloud/integration/integration-pytorch-torchtune.html)\n- [XTuner](https://docs.swanlab.cn/guide_cloud/integration/integration-xtuner.html)\n- [MMEngine](https://docs.swanlab.cn/guide_cloud/integration/integration-mmengine.html)\n- [FastAI](https://docs.swanlab.cn/guide_cloud/integration/integration-fastai.html)\n- [LightGBM](https://docs.swanlab.cn/guide_cloud/integration/integration-lightgbm.html)\n- [XGBoost](https://docs.swanlab.cn/guide_cloud/integration/integration-xgboost.html)\n\n**评估框架**\n- [EvalScope](https://docs.swanlab.cn/guide_cloud/integration/integration-evalscope.html)\n\n**计算机视觉**\n- [Ultralytics](https://docs.swanlab.cn/guide_cloud/integration/integration-ultralytics.html)\n- [MMDetection](https://docs.swanlab.cn/guide_cloud/integration/integration-mmdetection.html)\n- [MMSegmentation](https://docs.swanlab.cn/guide_cloud/integration/integration-mmsegmentation.html)\n- [PaddleDetection](https://docs.swanlab.cn/guide_cloud/integration/integration-paddledetection.html)\n- [PaddleYOLO](https://docs.swanlab.cn/guide_cloud/integration/integration-paddleyolo.html)\n\n**强化学习**\n- [Stable Baseline3](https://docs.swanlab.cn/guide_cloud/integration/integration-sb3.html)\n- [veRL](https://docs.swanlab.cn/guide_cloud/integration/integration-verl.html)\n- [HuggingFace trl](https://docs.swanlab.cn/guide_cloud/integration/integration-huggingface-trl.html)\n- [EasyR1](https://docs.swanlab.cn/guide_cloud/integration/integration-easyr1.html)\n\n**其他框架：**\n- [Tensorboard](https://docs.swanlab.cn/guide_cloud/integration/integration-tensorboard.html)\n- [Weights\u0026Biases](https://docs.swanlab.cn/guide_cloud/integration/integration-wandb.html)\n- [MLFlow](https://docs.swanlab.cn/guide_cloud/integration/integration-mlflow.html)\n- [HuggingFace Accelerate](https://docs.swanlab.cn/guide_cloud/integration/integration-huggingface-accelerate.html)\n- [Unsloth](https://docs.swanlab.cn/guide_cloud/integration/integration-unsloth.html)\n- [Hydra](https://docs.swanlab.cn/guide_cloud/integration/integration-hydra.html)\n- [Omegaconf](https://docs.swanlab.cn/guide_cloud/integration/integration-omegaconf.html)\n- [OpenAI](https://docs.swanlab.cn/guide_cloud/integration/integration-openai.html)\n- [ZhipuAI](https://docs.swanlab.cn/guide_cloud/integration/integration-zhipuai.html)\n\n[更多集成](https://docs.swanlab.cn/zh/guide_cloud/integration/integration-pytorch-lightning.html)\n\n\u003cbr\u003e\n\n## 🔌 插件\n\n欢迎通过插件来拓展SwanLab的功能，增强你的实验管理体验！\n\n- [自定义你的插件](https://docs.swanlab.cn/plugin/custom-plugin.html)\n- [邮件通知](https://docs.swanlab.cn/plugin/notification-email.html)\n- [飞书通知](https://docs.swanlab.cn/plugin/notification-lark.html)\n- [钉钉通知](https://docs.swanlab.cn/plugin/notification-dingtalk.html)\n- [企业微信通知](https://docs.swanlab.cn/plugin/notification-wxwork.html)\n- [Discord通知](https://docs.swanlab.cn/plugin/notification-discord.html)\n- [Slack通知](https://docs.swanlab.cn/plugin/notification-slack.html)\n- [CSV记录器](https://docs.swanlab.cn/plugin/writer-csv.html)\n\n\u003cbr\u003e\n\n## 🎮 硬件记录\n\nSwanLab会对AI训练过程中所使用的**硬件信息**和**资源使用情况**进行记录，下面是支持情况表格：\n\n| 硬件 | 信息记录 | 资源监控 | 脚本 |\n| --- | --- | --- | --- |\n| 英伟达GPU | ✅ | ✅ | [nvidia.py](https://github.com/SwanHubX/SwanLab/blob/main/swanlab/data/run/metadata/hardware/gpu/nvidia.py) |\n| 昇腾NPU | ✅ | ✅ | [ascend.py](https://github.com/SwanHubX/SwanLab/blob/main/swanlab/data/run/metadata/hardware/npu/ascend.py) |\n| 寒武纪MLU | ✅ | ✅ | [cambricon.py](https://github.com/SwanHubX/SwanLab/blob/main/swanlab/data/run/metadata/hardware/mlu/cambricon.py) |\n| CPU | ✅ | ✅ | [cpu.py](https://github.com/SwanHubX/SwanLab/blob/main/swanlab/data/run/metadata/hardware/cpu.py) |\n| 内存 | ✅ | ✅ | [memory.py](https://github.com/SwanHubX/SwanLab/blob/main/swanlab/data/run/metadata/hardware/memory.py) |\n| 硬盘 | ✅ | ✅ | [disk.py](https://github.com/SwanHubX/SwanLab/blob/main/swanlab/data/run/metadata/hardware/disk.py) |\n| 网络 | ✅ | ✅ | [network.py](https://github.com/SwanHubX/SwanLab/blob/main/swanlab/data/run/metadata/hardware/network.py) |\n\n如果你希望记录其他硬件，欢迎提交Issue与PR！\n\n\u003cbr\u003e\n\n## 🆚 与熟悉的工具的比较\n\n### Tensorboard vs SwanLab\n\n- **☁️ 支持在线使用**：\n  通过 SwanLab 可以方便地将训练实验在云端在线同步与保存，便于远程查看训练进展、管理历史项目、分享实验链接、发送实时消息通知、多端看实验等。而 Tensorboard 是一个离线的实验跟踪工具。\n\n- **👥 多人协作**：\n  在进行多人、跨团队的机器学习协作时，通过 SwanLab 可以轻松管理多人的训练项目、分享实验链接、跨空间交流讨论。而 Tensorboard 主要为个人设计，难以进行多人协作和分享实验。\n\n- **💻 持久、集中的仪表板**：\n  无论你在何处训练模型，无论是在本地计算机上、在实验室集群还是在公有云的 GPU 实例中，你的结果都会记录到同一个集中式仪表板中。而使用 TensorBoard 需要花费时间从不同的机器复制和管理\n  TFEvent 文件。\n\n- **💪 更强大的表格**：\n  通过 SwanLab 表格可以查看、搜索、过滤来自不同实验的结果，可以轻松查看数千个模型版本并找到适合不同任务的最佳性能模型。\n  TensorBoard 不适用于大型项目。\n\n### Weights and Biases vs SwanLab\n\n- Weights and Biases 是一个必须联网使用的闭源 MLOps 平台\n\n- SwanLab 不仅支持联网使用，也支持开源、免费、自托管的版本\n\n\u003cbr\u003e\n\n## 👥 社区\n\n### 周边仓库\n\n- [SwanLab-Docs](https://github.com/swanhubx/swanlab-docs)：官方文档仓库\n- [SwanLab-Dashboard](https://github.com/swanhubx/swanlab-dashboard)：离线看板仓库，存放了由`swanlab watch`打开的轻量离线看板的web代码\n- [self-hosted](https://github.com/swanhubx/self-hosted)：私有化部署脚本仓库\n\n### 社区与支持\n\n- [GitHub Issues](https://github.com/SwanHubX/SwanLab/issues)：使用 SwanLab 时遇到的错误和问题\n- [电子邮件支持](zeyi.lin@swanhub.co)：反馈关于使用 SwanLab 的问题\n- \u003ca href=\"https://docs.swanlab.cn/guide_cloud/community/online-support.html\"\u003e微信交流群\u003c/a\u003e：交流使用 SwanLab 的问题、分享最新的 AI 技术\n\n### SwanLab README 徽章\n\n如果你喜欢在工作中使用 SwanLab，请将 SwanLab 徽章添加到你的 README 中：\n\n[![][tracking-swanlab-shield]][tracking-swanlab-shield-link]、[![][visualize-swanlab-shield]][visualize-swanlab-shield-link]\n\n```\n[![](https://raw.githubusercontent.com/SwanHubX/assets/main/badge2.svg)](your experiment url)\n[![](https://raw.githubusercontent.com/SwanHubX/assets/main/badge1.svg)](your experiment url)\n```\n\n更多设计素材：[assets](https://github.com/SwanHubX/assets)\n\n### 在论文中引用 SwanLab\n\n如果您发现 SwanLab 对您的研究之旅有帮助，请考虑以下列格式引用：\n\n```bibtex\n@software{Zeyilin_SwanLab_2023,\n  author = {Zeyi Lin, Shaohong Chen, Kang Li, Qiushan Jiang, Zirui Cai,  Kaifang Ji and {The SwanLab team}},\n  doi = {10.5281/zenodo.11100550},\n  license = {Apache-2.0},\n  title = {{SwanLab}},\n  url = {https://github.com/swanhubx/swanlab},\n  year = {2023}\n}\n```\n\n### 为 SwanLab 做出贡献\n\n考虑为 SwanLab 做出贡献吗？首先，请花点时间阅读 [贡献指南](CONTRIBUTING.md)。\n\n同时，我们非常欢迎通过社交媒体、活动和会议的分享来支持 SwanLab，衷心感谢！\n\n\u003cbr\u003e\n\n**Contributors**\n\n\u003ca href=\"https://github.com/swanhubx/swanlab/graphs/contributors\"\u003e\n  \u003cimg src=\"https://contrib.rocks/image?repo=swanhubx/swanlab\" /\u003e\n\u003c/a\u003e\n\n\u003cbr\u003e\n\n## 📃 协议\n\n本仓库遵循 [Apache 2.0 License](https://github.com/SwanHubX/SwanLab/blob/main/LICENSE) 开源协议\n\n## Star History\n\n[![Star History Chart](https://api.star-history.com/svg?repos=swanhubx/swanlab\u0026type=Date)](https://star-history.com/#swanhubx/swanlab\u0026Date)\n\n\u003c!-- link --\u003e\n\n[release-shield]: https://img.shields.io/github/v/release/swanhubx/swanlab?color=369eff\u0026labelColor=black\u0026logo=github\u0026style=flat-square\n[release-link]: https://github.com/swanhubx/swanlab/releases\n\n[license-shield]: https://img.shields.io/badge/license-apache%202.0-white?labelColor=black\u0026style=flat-square\n[license-shield-link]: https://github.com/SwanHubX/SwanLab/blob/main/LICENSE\n\n[last-commit-shield]: https://img.shields.io/github/last-commit/swanhubx/swanlab?color=c4f042\u0026labelColor=black\u0026style=flat-square\n[last-commit-shield-link]: https://github.com/swanhubx/swanlab/commits/main\n\n[pypi-version-shield]: https://img.shields.io/pypi/v/swanlab?color=orange\u0026labelColor=black\u0026style=flat-square\n[pypi-version-shield-link]: https://pypi.org/project/swanlab/\n\n[pypi-downloads-shield]: https://static.pepy.tech/badge/swanlab?labelColor=black\u0026style=flat-square\n[pypi-downloads-shield-link]: https://pepy.tech/project/swanlab\n\n[swanlab-cloud-shield]: https://img.shields.io/badge/Product-SwanLab云端版-636a3f?labelColor=black\u0026style=flat-square\n[swanlab-cloud-shield-link]: https://swanlab.cn/\n\n[wechat-shield]: https://img.shields.io/badge/WeChat-微信-4cb55e?labelColor=black\u0026style=flat-square\n[wechat-shield-link]: https://docs.swanlab.cn/guide_cloud/community/online-support.html\n\n[colab-shield]: https://colab.research.google.com/assets/colab-badge.svg\n[colab-shield-link]: https://colab.research.google.com/drive/1RWsrY_1bS8ECzaHvYtLb_1eBkkdzekR3?usp=sharing\n\n[github-stars-shield]: https://img.shields.io/github/stars/swanhubx/swanlab?labelColor\u0026style=flat-square\u0026color=ffcb47\n[github-stars-link]: https://github.com/swanhubx/swanlab\n\n[github-issues-shield]: https://img.shields.io/github/issues/swanhubx/swanlab?labelColor=black\u0026style=flat-square\u0026color=ff80eb\n[github-issues-shield-link]: https://github.com/swanhubx/swanlab/issues\n\n[github-contributors-shield]: https://img.shields.io/github/contributors/swanhubx/swanlab?color=c4f042\u0026labelColor=black\u0026style=flat-square\n[github-contributors-link]: https://github.com/swanhubx/swanlab/graphs/contributors\n\n[demo-cats-dogs]: https://swanlab.cn/@ZeyiLin/Cats_Dogs_Classification/runs/jzo93k112f15pmx14vtxf/chart\n[demo-cats-dogs-image]: readme_files/example-catsdogs.png\n\n[demo-yolo]: https://swanlab.cn/@ZeyiLin/ultratest/runs/yux7vclmsmmsar9ear7u5/chart\n[demo-yolo-image]: readme_files/example-yolo.png\n\n[demo-qwen2-sft]: https://swanlab.cn/@ZeyiLin/Qwen2-fintune/runs/cfg5f8dzkp6vouxzaxlx6/chart\n[demo-qwen2-sft-image]: readme_files/example-qwen2.png\n\n[demo-google-stock]:https://swanlab.cn/@ZeyiLin/Google-Stock-Prediction/charts\n[demo-google-stock-image]: readme_files/example-lstm.png\n\n[demo-audio-classification]:https://swanlab.cn/@ZeyiLin/PyTorch_Audio_Classification/charts\n[demo-audio-classification-image]: readme_files/example-audio-classification.png\n\n[demo-qwen2-vl]:https://swanlab.cn/@ZeyiLin/Qwen2-VL-finetune/runs/pkgest5xhdn3ukpdy6kv5/chart\n[demo-qwen2-vl-image]: readme_files/example-qwen2-vl.jpg\n\n[demo-easyr1-rl]:https://swanlab.cn/@Kedreamix/easy_r1/runs/wzezd8q36bb6dlza6wtpc/chart\n[demo-easyr1-rl-image]: readme_files/example-easyr1-rl.png\n\n[demo-qwen2-grpo]:https://swanlab.cn/@kmno4/Qwen-R1/runs/t0zr3ak5r7188mjbjgdsc/chart\n[demo-qwen2-grpo-image]: readme_files/example-qwen2-grpo.png\n\n[tracking-swanlab-shield-link]:https://swanlab.cn\n[tracking-swanlab-shield]: https://raw.githubusercontent.com/SwanHubX/assets/main/badge2.svg\n\n[visualize-swanlab-shield-link]:https://swanlab.cn\n[visualize-swanlab-shield]: https://raw.githubusercontent.com/SwanHubX/assets/main/badge1.svg\n\n[dockerhub-shield]: https://img.shields.io/docker/v/swanlab/swanlab-next?color=369eff\u0026label=docker\u0026labelColor=black\u0026logoColor=white\u0026style=flat-square\n[dockerhub-link]: https://hub.docker.com/r/swanlab/swanlab-next/tags","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fswanhubx%2Fswanlab","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fswanhubx%2Fswanlab","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fswanhubx%2Fswanlab/lists"}